Zhang Yiming
Zhang Yiming is a Chinese entrepreneur who founded ByteDance in 2012 and built it into the owner of TikTok, Douyin and Toutiao, among the most widely used applications in the world. He stepped down as chief executive in 2021 and as chairman later that year, and has since been reported as the wealthiest person in China.
ByteDance's significance rests on a single technical idea applied consistently: that content should be distributed by a recommendation system inferring what a user wants from their behaviour, rather than by the user choosing whom to follow. That approach, first applied to news aggregation and then to short video, changed how social media works worldwide and forced every major competitor to rebuild its products around it.
Early life and education
Zhang was born on 1 April 1983 in Longyan, a city in Fujian province in south-eastern China. His parents were civil servants.
He studied at Nankai University in Tianjin, beginning in microelectronics and transferring to software engineering, and graduated in 2005.
Career
Early ventures
Zhang's first job was at Kuxun, a travel search start-up, where he became responsible for engineering within a short period. He then spent a brief period at Microsoft, which he has said he found too constrained, and in 2009 founded 99fang.com, a property search site.
The pattern of these ventures anticipated what followed: all were concerned with helping people find information in a market where the volume of it exceeded anyone's ability to search it manually.
ByteDance
Toutiao
Zhang founded ByteDance in March 2012 in a Beijing apartment. Its first significant product was Jinri Toutiao, meaning Today's Headlines, a news aggregation application launched in August 2012.
Toutiao's distinguishing feature was that it had no editors and required no choices from the user. Rather than presenting a publication's front page or a list of subscriptions, it observed what each user read, how long they spent, what they skipped, and assembled a personalised feed accordingly. The system improved with use and required no effort from the reader, and within a few years the application had hundreds of millions of users.
The approach was contentious in China, where news distribution is closely regulated. Regulators objected repeatedly to the content the algorithm promoted, and in April 2018 the authorities ordered the permanent closure of Neihan Duanzi, a ByteDance application for humour and short video that had a very large following. Zhang published a public apology stating that the product had deviated from core socialist values and that the company had placed insufficient weight on guidance, and committed to expanding its content moderation workforce substantially. The episode is the clearest illustration of the constraints under which the company operates domestically.
Douyin and TikTok
Douyin was launched in China in September 2016, applying the same recommendation approach to short vertical video. TikTok was launched internationally in 2017 as a separate application operating on the same technology but with separate infrastructure and content.
The decisive transaction was the acquisition of Musical.ly in late 2017 for a sum reported at between US$800 million and US$1 billion. Musical.ly was a Shanghai-based company whose lip-syncing application had become popular among teenagers in the United States and Europe, a market Chinese consumer applications had never previously reached at scale. ByteDance merged Musical.ly into TikTok in August 2018, transferring its users onto the superior recommendation technology.
The result was the first Chinese consumer internet product to achieve mass adoption in Western markets. TikTok passed a billion monthly users within a few years and became the most downloaded application in the world for several consecutive years.
The mechanism
TikTok's recommendation system differs from its predecessors in a way worth stating precisely, because it explains both the product's success and the competitive response.
Conventional social media distributes content through a social graph: a user sees what the people they have chosen to follow have posted. This makes the product's value dependent on the user having built a network, which is slow, and it limits what can be shown to what that network happens to produce.
TikTok distributes through an interest graph inferred from behaviour. A new user with no connections receives a sequence of videos and the system learns from what they watch, rewatch, skip and share. A video from an account with no followers can reach millions of people if the system's initial tests indicate that viewers respond to it. This makes the product immediately useful to a new user and gives any creator a genuine chance of reaching an audience.
Every large competitor rebuilt around the same principle. Meta Platforms introduced Reels across Instagram and Facebook and shifted its main feeds towards recommended rather than followed content; Google introduced YouTube Shorts; Snapchat and others followed. The industry's convergence on ByteDance's model within a few years is the clearest measure of the idea's force.
Stepping back
Zhang announced in May 2021 that he would step down as chief executive, with effect from the end of the year, and was succeeded by Liang Rubo, a university contemporary and co-founder. He relinquished the chairmanship in November 2021 and has since held no executive position, retaining a substantial shareholding and a majority of the voting rights.
The stated reason was that he was more interested in research and long-range thinking than in management, and that he lacked the social qualities the role required. The timing coincided with an intense period of regulatory action against Chinese technology companies, and a number of other founders, including those of Pinduoduo and JD.com, stepped back in the same period, which has led to persistent speculation that the decisions were not purely personal.
The TikTok divestiture
TikTok's ownership became a matter of national security policy in the United States and elsewhere.
The concerns advanced were twofold: that data on American users could be obtained by the Chinese government under laws requiring companies to assist state intelligence work, and that the recommendation system could be used to influence what a very large proportion of the American population sees. TikTok's response was Project Texas, an arrangement under which American user data was held by Oracle on American infrastructure and source code was made available for inspection, which the company argued addressed the concerns and which critics argued did not address control of the algorithm.
An attempt to force a sale in 2020 by executive order was blocked in the courts. Legislation passed in April 2024, the Protecting Americans from Foreign Adversary Controlled Applications Act, required divestiture of TikTok's American operations or prohibited the application. The Supreme Court upheld the statute in January 2025, and the application was briefly unavailable in the United States before enforcement was deferred.
Arrangements were subsequently negotiated under which TikTok's American operations would pass to a joint venture with substantial American ownership, with Oracle, Silver Lake and MGX among the investors, and with ByteDance retaining a minority interest. The treatment of the recommendation algorithm, which the Chinese government had in 2020 added to its export control list specifically to give itself a veto over any transfer, was the most contested element of the negotiation.
Wealth and public profile
Zhang has been reported as the wealthiest individual in China, his fortune derived from a holding in ByteDance of roughly a fifth, valued through periodic share repurchases the company conducts from employees and investors, which have implied valuations of several hundred billion dollars.
He is notably private, gives few interviews, and has maintained a lower public profile than any comparable founder. He has been reported as living principally in Singapore.
Assessment
Zhang built, in under a decade, the only Chinese consumer internet company to succeed at scale outside China, and did so with a product idea that the entire industry subsequently adopted.
The technical insight was genuine and was not obvious. That people would rather be shown what they will like than choose it themselves runs against the assumption underlying twenty years of social media design, and ByteDance demonstrated it first in news, then in humour, then in video, in each case against incumbents with far greater resources.
The consequences are more contested than the achievement. A distribution system optimised for engagement promotes what holds attention, which is not the same as what is true, useful or good for the person watching, and the literature on the effects of short video on attention, sleep and adolescent mental health has grown steadily. The system is also opaque: neither users nor regulators can readily establish why they are shown what they are shown, which is the substance of the concern about political influence.
Zhang's own position is the more unusual part. He built the company, established its method, stepped away from it at thirty-eight, and has said almost nothing publicly since. Whether that reflects the preference he stated or the conditions facing prominent Chinese entrepreneurs after 2021 is not determinable from outside, and both explanations are consistent with what is known.